Академический Документы
Профессиональный Документы
Культура Документы
60 (2009) 899–920
0044-2275/09/050899-22
DOI 10.1007/s00033-009-8056-z Zeitschrift für angewandte
°c 2009 Birkhäuser Verlag, Basel Mathematik und Physik ZAMP
Suvadip Paul
Abstract. A method of moment is employed to study the axial dispersion of passive tracer
molecules released in an unsteady pressure-driven flow through an annular pipe which is oscil-
lating around its longitudinal axis. The flow unsteadiness is caused by the oscillation of the tube
around its axis as well as by a periodic pressure gradient. A finite difference implicit scheme
is adopted to solve the Aris integral moment equations arising from the unsteady convective-
diffusion equation for all time periods. The main objective is to study the nature of the dispersion
coefficient and mean concentration distribution under the sole as well as combined oscillation of
the two driving forces. The behaviour of the dispersion coefficient due to the variation of the
aspect ratio, the absorption parameter for purely periodic flow has been examined and the sound
response from dispersion coefficient is found with the variation of these parameters in the sole
presence of pressure pulsation. There is a remarkable difference in the behavior of the disper-
sion coefficient depending on whether the ratio of two frequencies arising from the oscillations
of the tube and the pressure gradient possesses a proper fraction or not. Oscillation of the tube
produces much more dispersion than the pulsation of the pressure gradient and their combined
effect leads to a further increase in dispersion. Tube oscillation shows a stronger effect on the
dispersion coefficient than the pressure pulsation though the effect of physical parameters are
pronounced in the presence of pressure pulsation.
The effect of the frequency parameter on the axial distribution of mean concentration is
insensible when the oscillation of the annular tube is the only forcing. However this effect is
much noticeable under the combined action of both forcing and much more effective under the
sole influence of pressure pulsation.
Keywords. Axial dispersion, finite difference, absorption, aspect ratio, concentration distribu-
tion, pressure gradient.
1. Introduction
2. Flow equations
3. Convection-diffusion equation
Suppose a finite cloud of passive solute is discharged into the flow at an arbitrary
instant in presence of first order reaction at the outer wall. Then the concentration
C(t, ρ, ζ) of the solute as a function of axial distance ζ, radial distance ρ and time
t satisfies the non-dimensional convective-diffusion equation,
µ ¶
∂C ∂C 1 ∂ ∂C ∂2C
+ Pe u(ρ, t) = ρ + (4)
∂t ∂ζ ρ ∂ρ ∂ρ ∂ζ 2
902 S. Paul ZAMP
4. Velocity distribution
In order to solve the equation (1) with the equation (2), we assume a solution of
the form as
Figure 1. Profiles of the unsteady component of velocity for different phase angle of pressure
pulsation when the phase angle Ω of the tube oscillation is π/2 ; ²1 = ²2 = 1.
tube oscillation, the tube will change its position along its axis. Fig. 1 shows the
velocity profiles for different phase angles of pressure pulsation when the annular
tube oscillates with the phase angle Ω = α2 St = π/2. The dotted lines represent
the case when there is no oscillation in the pressure gradient. From the figure it is
clear that how the term α1 St that characterizes the pressure pulsation, influences
the velocity distribution. The phase angle Ω of the tube oscillation is also found
to be influential. It has seen that the velocity increases as the phase angle Ω of
tube oscillation increases. At a particular phase angle (of tube oscillation), the
velocity decreases with the increase of the phase angle of the pressure pulsation.
Under the sole influence of the oscillation of the pressure gradient, as the fre-
quency parameter α1 increases, the velocity over the annular gap of the tube
decreases. It happens that at a given radial position velocity decreases with the
increase of aspect ratio (figure is not provided).
5. Moment equations
Following the method of integral moment proposed by Aris [2], we define the p-th
moment of the distribution of the solute in the filament through ρ at time t as,
Z +∞
Cp (t, ρ) = ζ p C(t, ρ, ζ)dζ (10)
−∞
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 905
Figure 2. Variation of the dispersion coefficient for steady flow with variation of F , λ and β.
and the concentration distribution of the solute over the cross-section of the an-
nulus is given by, R 2π R 1
dθ ρCp (t, ρ)dρ
C̄p = 0 R 2π λ R 1 . (11)
0
dθ λ ρdρ
So using equations (10), the diffusion equation (4) subject to initial and bound-
ary conditions can be written as
µ ¶
∂Cp 1 ∂ ∂Cp
− ρ = pu(ρ, t)Pe Cp−1 + p(p − 1)Cp−2 (12)
∂t ρ ∂ρ ∂ρ
with
½
B(ρ) for p = 0
Cp (0, ρ) =
0 for p > 0
∂Cp
+ βCp = 0 at ρ = 1, (13)
∂ρ
∂Cp
= 0 at ρ = λ
∂ρ
where B(ρ) = 1 for (λ ≤ ρ ≤ 1).
906 S. Paul ZAMP
Averaging over the annular cross-section, equations (12) and (13) become,
dC̄p 2β
= pPe u(ρ, t)C p−1 + p(p − 1)C p−2 − Cp (t, 1) (14)
dt 1 − λ2
and
C̄p (0) = 1 for p = 0, C̄p (0) = 0 for p>0 (15)
where over-bar denotes the cross-sectional mean.
The p-th order central moment of the concentration distribution about the
mean can be defined as
1
R 1 R 2π R +∞
π λ 0 −∞
ρ(ζ − ζg )p Cdρdθdζ
µp (t) = RRR (16)
Cdv
C̄1
where ζg = is the centroid or first moment of the solute and C̄0 represents the
C̄0
total mass of the reactive solute in the whole volume of the annular pipe.
Each of the integral moments of concentration defined by (16) has an impor-
tant contribution for predicting dispersion phenomena. They serve as simple and
physically meaningful descriptors of the overall behaviour of the slug. The first
moment ζg measures the location of the center of gravity of the slug movement
with the mean velocity of the fluid, initially located at the source. The second
central moment µ2 represents the variance related to the dispersion of the contam-
inant about the mean position. The skewness factor (ν2 ) and the flatness factor
(ν3 ), obtained from the third and the fourth-order moments, are also important
factors during the initial stage of matter dispersion.
6. Numerical solution
As the analytical solution of moment equations (12) and (14) subject to the initial
and boundary conditions (13), (15) is somewhat complicated, a finite difference
method based on the Crank–Nicholson implicit scheme has been adopted to study
the dispersion phenomenon. The equations are discretized in the radial direction
ρ and time t. For this purpose, we divide the whole width of the annulus into
(M − 1) equal parts each of length ∆ρ which are represented by the grid point j,
so that j = 1, M corresponds to the inner wall ρ = λ and outer wall ρ = 1 of the
annulus respectively i.e., ρj = λ + (j − 1) × ∆ρ. The grid point i identifies time
t according to the relationship ti = ∆t × (i − 1) so that i = 1 corresponds to the
time t = 0. The Crank-Nicholson method is applied for each time step. ∆t and
∆ρ are the increments of t and ρ, respectively. Cp (i, j) indicates the value of Cp at
the i-th grid point along the t-axis and j-th grid point along ρ-axis. The resulting
finite difference equations become a system of linear algebraic equations with a
tri-diagonal coefficient matrix and can be represented in the form,
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 907
Pj Cp (i + 1, j + 1) + Qj Cp (i + 1, j) + Rj Cp (i + 1, j − 1) = Sj (17)
A matlab code has been developed to solve this tri-diagonal coefficient matrix
by the method of Thomas algorithm [Anderson et al. [17]] with the help of pre-
scribed initial and boundary conditions. The steps of computations are: (i) first
the time dependent axial velocity u is computed from (6); (ii) the concentration Cp
is calculated from (12) on knowing the values of u(ρ, t) at the grid point (i + 1, j)
and (iii) finally the spatial moments C̄p are calculated from (14) by applying Simp-
son’s one-third rule, with the known values of u(ρ, t) and Cp at the corresponding
grid points. The values of the variables can be calculated for all time iteratively
in the marching direction. Although the present scheme is linearly stable for any
∆t 1−λ
finite values of 2
, a mesh size (∆t = 0.0001, ∆ρ = ) gives satisfactory
(∆ρ) M −1
results for steady currents. For the oscillatory and combined flow, a mesh size
1−λ
(∆t = 0.00001, ∆ρ = ) where λ varies from 0.02 to 0.2, gives acceptable
M −1
results for the frequencies α1 or α2 = 0.5, 1.0 and 2.0. Actually, the smaller time
interval is needed to trap the oscillatory nature in the dispersion process.
Knowledge of the various moments of the material concentration suggests the be-
haviour of the concentration distribution. Using the first four central moments, it
is possible to compute the mean axial concentration distribution Cm (ζ, t) of trac-
ers within the annular region with the help of Hermite polynomial representation
for non-Gaussian curves (Chatwin [18]) and is given by,
∞
X
2
Cm (ζ, t) = C̄0 (t)e−z an (t)Hn (z) (18)
n=0
908 S. Paul ZAMP
ζ − ζg C̄1
where z = 1/2
, ζg =and Hi , the Hermite polynomials, satisfy the
(2µ2 ) C̄0
recurrence relation with H0 (z) = 1.0 as
Hi+1 (z) = 2zHi (z) − 2iHi−1 (z), i = 0, 1, 2, . . .
The coefficients ai are
1/2
a0 = 1/ (2πµ2 ) , a1 = a2 = 0, a3 = 21/2 a0 ν2 /24, a4 = a0 ν3 /96.
Therefore, given the statistical parameters (16), the concentration distribution
can be estimated from (18) at any given location in the axial direction and time.
When the flow is purely oscillatory, the substitution of ²2 = 0 will make the fluid to
flow through a stationary annular tube driven by a periodic pressure gradient. The
behaviour of the dispersion coefficient in this type of flow situation is in complete
agreement with Mazumder and Mondal [10].
Also to defend the results obtained, the same problem has also been solved using
the derivative expansion method developed by Gill and Sankarasubramanian [19].
Following the approach of Sarkar and Jayaraman [20] the absorption coefficient
M0 (t), convective coefficient M1 (t) and dispersion coefficient M2 (t) is determined
(space details are omitted). The dispersion coefficient M2 (t) obtained in this
way and the dispersion coefficient Da obtained using Aris method show similar
qualitative behaviour.
The second spatial moment corresponds to the variance of the concentration
distribution, the rate of change of which gives the dispersion coefficient. The rate
of growth of variance which indicates the degree of the dispersion effect at any
time is given by
1 dµ2
Da = (19)
2Pe2 dt
where Da is the apparent dispersion coefficient depending on the frequencies of
oscillations αi , amplitudes of oscillations ²i (i = 1, 2), Schmidt number S, absorp-
tion parameter β, aspect ratio λ, applied pressure gradient F and the dispersion
time t.
When the flow is steady (u = u0 (ρ)), the apparent dispersion coefficient Da
initially increases with time and asymptotically reaches a stationary state. It is
seen from Fig. 2 that Da decreases with the increase in absorption parameter β.
For a given instant of time t, Da decreases with F , but increases as λ decreases.
Variation of the dispersion coefficient against time under the sole influence of
only one forcing is presented in Fig. 3. For Fig. 3(a, b, c, d), the pressure gradient
is the only driving force (i.e., ²1 6= 0, ²2 = 0) while for Fig. 4(a, b, c, d) it is the
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 909
Figure 3. The dispersion coefficient Da for different frequency α1 of pressure pulsation in the
absence of tube oscillation (²2 =0) when F = 1, ²1 = 1, Pe =S=103 .
oscillation of the annular tube around its axis (i.e., ²1 = 0, ²2 6= 0). At a given
frequency, the dispersion coefficient shows more variability under the oscillation
of the annular tube compared with the oscillation of the pressure gradient. It is
interesting to note that the oscillation of the annular tube produces more dispersion
than that due to the periodic pressure gradient and in both cases increase of
frequency leads to a reduction of the dispersion coefficient. This is because the
increase of the frequency parameter decreases the flow within the annular region
and the aggregation of tracer molecules comes to appear in a smaller portion in
the axial direction. Fig. 3(d) or 4(d) correspond to the case when the pressure
gradient or the annular tube oscillates with sufficiently high frequency (O(10)), a
case which gives rise to thinner Stoke’s layers. From Fig. 3 it is clear that, for low
frequency of pressure pulsation the dispersion coefficient Da changes cyclically with
a double-frequency period during initial stage. The amplitudes of oscillation of Da
during the first and second half of the period of oscillatory flow are approximately
910 S. Paul ZAMP
Figure 4. The dispersion coefficient Da for different frequency α2 of tube oscillation in the
absence of pressure pulsation (²1 =0) when F = 1, ²2 =1, Pe =S=103 .
equal at low frequency, but in the case of high frequency [Fig. 3(d)], the period is
so short that Da initially varies with almost single-frequency oscillation. More or
less similar behaviour can be seen from Fig. 4 which corresponds to the oscillation
of the annular tube.
For oscillatory current, the variations of Da with respect to time are depicted
for different values of β in Fig. 5 for λ = 0.1 and for different values of λ in Fig. 6
for β = 0 when (i) the pressure gradient oscillates with frequency α1 = 0.5 but
the annular tube is stationary (i.e., ²2 = 0), (ii) the annular tube oscillates with
frequency α2 = 0.5 but there is no oscillation in the pressure gradient (i.e., ²1 = 0)
and (iii) the pressure gradient and the annular tube both oscillate with frequency
0.5 i.e., α1 = α2 = 0.5. It is seen from Fig. 5 that in all cases increase in absorption
at the boundary helps to decrease the dispersion coefficient. The scale of reduction
is negligible when tube oscillation is the only driving force [Fig. 5(c)]. But under
the sole influence of pressure gradient oscillation and under the combined action
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 911
Figure 5. Temporal variation of dispersion coefficient Da for various values of β under different
types of unsteady flow situations where unsteadiness is caused by the (a) combined action of
both the driving forces (²1 = ²2 = 1) (b) sole oscillation of the pressure gradient
(²1 = 1, ²2 = 0) and (c) sole oscillation of the annular tube (²1 = 0, ²2 = 1).
of both of the driving forces, the dispersion coefficient shows appreciable lowering
with the increase of absorption in the boundary [Fig. 5(a, b)]. Thus presence of
pressure pulsation helps Da to reduce with absorption on a larger scale.
Increase of aspect ratio also leads to a decrease of the dispersion coefficient
[Fig. 6]. This can be explained by the fact that the increase of λ leads to a
decrease of the flow within the annular region which causes a lower dispersion
rate. In the sole presence of pressure pulsation or in the presence of both of the
driving forces, there is considerable reduction in Da with the increase of aspect
ratio. If oscillation of the tube is the only driving force [Fig. 6(c)], then also the
912 S. Paul ZAMP
dispersion coefficient reduces with the increase of the aspect ratio, but the scale
of reduction is so small that it seems that the dispersion coefficient is insensible
with the variation of the aspect ratio, in the absence of pressure pulsation.
In order to study the simultaneous effect of pressure pulsation and tube oscil-
lation on dispersion, non-zero values of ²1 and ²2 are taken. For the oscillatory
flow the variations of Da w.r.t time are depicted for different values of α1 and
α2 for small time in Fig. 7 and for large time in Fig. 8. The figures along the
horizontal direction (row-wise) explain how the dispersion coefficient changes with
the gradual increase in α2 when α1 is fixed at different level and the figures ar-
ranged in the vertical direction (column-wise) describe the change in dispersion
coefficient Da with α1 when α2 is fixed at different level. Comparison of the rows
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 913
Figure 7. Dispersion coefficient Da with time for different values of α1 and α2 during initial
stage when ²1 = ²2 =1, F = 1, Pe =S=103 : (a, e, i) for α1 =α2 ; (d, g, h) for α1 > α2 ; (b, c, f)
for α1 < α2 .
or columns reveals the change in the dispersion coefficient with the changes in α1
and α2 . Diagonal plots (a, e, i) of Figs. 7 and 8 correspond to α1 = α2 . i.e.,
the pressure gradient and the annular tube oscillate with the same frequency. It is
seen from Figs. 7(a, e, i) that in periodic flow the dispersion coefficient Da changes
cyclically with a double-frequency period during initial stage and it reaches a sta-
tionary state after a certain time [Fig. 8(a, e, i)]. Figures show that with the
increase of α1 and α2 , the fluctuation in Da increases and reaches a stationary
state later than the smaller frequency. It should be noted that for all time the
increase of either frequency leads to a reduction of dispersion coefficient Da . The
plots (d, g, h) and (b, c, f) in Figs. 7 and 8 describe the cases when α1 > α2
914 S. Paul ZAMP
Figure 8. Dispersion coefficient Da at large time for different values of α1 and α2 when
²1 = ²2 =1, F = 1, Pe =S=103 : (a, e, i) for α1 =α2 ; (d, g, h) for α1 > α2 ; (b, c, f) for α1 < α2 .
to a further increase in dispersion. From the figures it is clear that during their
combined action the oscillation of the annular tube plays a more important role
on Da than the pulsation of the pressure gradient.
Fig. 10 plots dispersion coefficient against time with the variation of amplitudes
(²1 and ²2 ) of the oscillations and it is observed from Fig. 10(a, b) that at a fixed
instant, the amplitude of Da increases with increase in both ²1 and ²2 , but the
scale of increment for ²1 is much less than that for ²2 . It should be noted that the
variation of the amplitude of tube oscillation has a stronger effect on dispersion
compared to that for pressure pulsation. It is interesting to note from Fig. 10(c)
916 S. Paul ZAMP
Figure 10. Variation of dispersion coefficient Da with time for F =1, α1 = α2 = 0.5,
P e = S = 103 : (a) for different values of ²1 when ²2 =1; (b) for different values of ²2 when
²1 =1, (c) when ²1 =²2 .
that if the pressure gradient and the annular tube oscillate with the same amplitude
(i.e., ²1 =²2 ), the increase of ²1 (=²2 ) has the same effect on Da as the increase of
²2 keeping ²1 fixed (Fig. 10(b)) proving again that oscillation of the annular tube
has more influence on Da than the pulsation of the pressure gradient.
The axial distribution of the mean concentration are approximated using Her-
mite polynomial representation from the first four central moments for a range
of different aspect ratios, absorption parameters and frequencies of oscillations.
The variations of mean concentration distribution Cm (ζ, t) have been presented in
Fig. 11 against the axial distance (ζ − ζg ). It is seen from the Fig. 11(a) that as
the dispersion time t increases, the peak of the distribution decreases and tends
to become flat (solid line). It is also clear from the figure (dotted line) that for
Vol. 60 (2009) Axial dispersion in pressure perturbed flow through an annular pipe 917
Figure 11. Axial distribution of the mean concentration Cm (ζ, t) for periodic flow when ²=1,
Pe = S = 1000: (a) for different dispersion time t (solid lines) and for different β (dotted lines);
(b) for different λ (dotted lines) and F (solid lines) and (c) for different α1 (dash-dotted lines),
for different α2 (solid lines) and for different but equal values of both α1 and α2 (dashed lines).
t = 0.2, λ = 0.1, the increase of reaction parameter β shows the depletion of the
amount of the reactive material, and therefore the peak of the mean concentration
distribution gradually decreases. It can be seen from Fig. 11(b) that the peak of
the concentration distribution decreases with the decrease in λ. Also an increase
in pressure gradient leads to a decrease of the peak of the mean concentration dis-
tribution. Fig. 11(c) shows that for t = 0.2, λ = 0.1, β = 1.0, F = 4, the peak of
the axial distribution of mean concentration increases with increase in frequency
parameters αi ’s. The effect of frequency parameter on mean concentration distri-
bution is insensible when the oscillation of the tube is the only forcing, though
918 S. Paul ZAMP
the effect is detectable under the combined action of both forcing and much more
pronounced under the sole influence of pressure pulsation. It should be mentioned
here that for relatively low values of the dimensionless mean pressure gradient,
increase of frequency parameter does not show any appreciable change in the
mean concentration distribution in all cases. Therefore as F decreases, the effect
of frequency parameters tends to become insignificant in the mean concentration
distribution. Another important finding is that sole oscillation of pressure gradient
produces more peaky distribution than that produced by the tube-oscillation and
the peak further decreases when both of the driving forces combines.
9. Conclusions
[6] Under the sole presence of tube oscillation mean concentration distribution is
unaffected with the variation of the frequency parameter. However this effect
is much detectable under the combined action of both forcing and much more
visible under the isolated influence of pressure pulsation.
Acknowledgement
References
[1] Taylor, G. I., Dispersion of soluble matter in solvent flowing slowly through a tube, Proc.
Roy. Soc. Lond. A 219 (1953), 186–203.
[2] Aris, R., On the dispersion of a solute in a fluid flowing through a tube, Proc. Roy. Soc.
Lond. A 235 (1956), 67–77.
[3] Barton, N. G., On the method of moments for solute dispersion, J. Fluid Mech. 126 (1983),
205–218.
[4] Aris, R., On the dispersion of a solute in a pulsating flow through a tube, Proc. Roy. Soc.
Lond. A 259 (1960), 370–376.
[5] Watson, E. J., Diffusion in oscillatory pipe flow, J. Fluid Mech. 133 (1983), 233–244.
[6] Tsangaris, S. and Athanassiadis, N., Diffusion in oscillatory flow in an annular pipe, ZAMM
65(4) (1985), T252–T254.
[7] Mazumder, B. S. and Das, S. K., Effect of boundary reaction on solute dispersion in pulsatile
flow through a tube, J. Fluid Mech. 239 (1992), 523–549.
[8] Mondal, K. K. and Mazumder, B. S., On the solute dispersion in a pipe of annular cross
section with absorbing boundary, ZAMM 85(6) (2005), 422-430.
[9] Sarkar, A. and Jayaraman, G., Non-linear analysis of oscillatory flow in the annulus of an
elastic tube - application to catheterized artery, Phys. of Fluids 13(10) (2001), 2901–2911.
[10] Mazumder, B. S. and Mondal, K. K., On solute transport in oscillatory flow through an
annular pipe with reactive wall and application to a catheterized artery, Quart. Jour. of
Mech. and Appl. Math. 58(3) (2005), 349–365.
[11] Waters, S. L., Solute uptake through the walls of a pulsating channel, J. Fluid Mech. 433
(2001), 193–208.
[12] Hydon, P. E. and Pedley, T. J., Axial dispersion in a channel with oscillating walls, J. Fluid
Mech. 249 (1993), 535–555.
[13] Secomb, T. W., Flow in a channel with pulsating walls, J. Fluid Mech. 88 (1978), 273–288.
[14] Ng, C. O. and Bai, Y. C., Dispersion in oscillatory Couette flow with sorptive boundaries,
Acta Mechanica 178 (2005), 65–84.
[15] Bandyopadhyay, S. and Mazumder, B. S., On contaminant dispersion in unsteady generalised
Couette flow, Int. J. Engg. Sci. 37 (1999), 1407–1423.
[16] Mazumder, B. S. and Paul, S., Dispersion in oscillatory Couette flow with absorbing bound-
aries, Int. J. Fluid Mech. Res. 35(5) (2008), 475–492.
[17] Anderson, D. A., Tanehill, J. C. and Pletcher, R. H., Computational Fluid Mechanics and
Heat Transfer, Hemisphere Publishing Corporation, New-York 1984.
920 S. Paul ZAMP
[18] Chatwin, P. C., The approach to normality of the concentration distribution of solute in a
solvent flowing along a straight tube, J. Fluid Mech. 43 (1970), 321–352.
[19] Gill, W. N. and Sankarsubramanian, R., Dispersion of a non-uniform slug in time dependent
flow, Proc. Roy. Soc. Lond. A 322 (1971), 101–117.
[20] Sarkar, A. and Jayaraman, G., The effect of wall absorption on dispersion in oscillatory flow
in an annulus: application to a catheterized artery, Acta Mech. 172 (2004), 151–167.
Suvadip Paul
Physics and Applied Mathematics Unit
Indian Statistical Institute
Kolkata 700 108
India
e-mail: paul suvadip@rediffmail.com